# The Alignment Floor: How Persona Customization Breaks Safety in Weakly-Aligned LLMs

arXiv: https://arxiv.org/abs/2605.27382 · HTML: https://mlsanity.com/paper/2605.27382/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in Human values and AI feedback alignment, one of 10 directions in LLM alignment / RLHF / preference optimisation.
- Builds on 15 of its 23 references that are in the map.
- First posted on arXiv 10 Apr 2026.

## Placed in

- LLM alignment / RLHF / preference optimisation / [Human values and AI feedback alignment](https://mlsanity.com/alignment/human-values-and-ai-feedback-alignment/)

## Builds on (most cited first, 15 of 15)

- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [Training Verifiers to Solve Math Word Problems](https://mlsanity.com/paper/2110.14168/) (2021)
- [Constitutional AI: Harmlessness from AI Feedback](https://mlsanity.com/paper/2212.08073/) (2022)
- [TruthfulQA: Measuring How Models Mimic Human Falsehoods](https://mlsanity.com/paper/2109.07958/) (2021)
- [Towards Understanding Sycophancy in Language Models](https://mlsanity.com/paper/2310.13548/) (2023)
- [Discovering Language Model Behaviors with Model-Written Evaluations](https://mlsanity.com/paper/2212.09251/) (2022)
- [Adversarial NLI: A New Benchmark for Natural Language Understanding](https://mlsanity.com/paper/1910.14599/) (2019)
- [Foundational Challenges in Assuring Alignment and Safety of Large Language Models](https://mlsanity.com/paper/2404.09932/) (2024)
- [OctoPack: Instruction Tuning Code Large Language Models](https://mlsanity.com/paper/2308.07124/) (2023)
- [Simple synthetic data reduces sycophancy in large language models](https://mlsanity.com/paper/2308.03958/) (2023)
- [The PRISM Alignment Dataset: What Participatory, Representative and Individualised Human Feedback Reveals About the Subjective and Multicultural Alignment of Large Language Models](https://mlsanity.com/paper/2404.16019/) (2024)
- [When Large Language Models contradict humans? Large Language Models' Sycophantic Behaviour](https://mlsanity.com/paper/2311.09410/) (2023)
- [The Power of Personality: A Human Simulation Perspective to Investigate Large Language Model Agents](https://mlsanity.com/paper/2502.20859/) (2025)
- [Behavioral Fingerprinting of Large Language Models](https://mlsanity.com/paper/2509.04504/) (2025)
- [Personality as a Probe for LLM Evaluation: Method Trade-offs and Downstream Effects](https://mlsanity.com/paper/2509.04794/) (2025)

## Built on by (most cited first, 1 of 1)

- [Beyond the Sycophancy Score: How Task, Model, and Pressure Shape LLM Yielding](https://mlsanity.com/paper/2610.08840/) (2026)

## Abstract (arXiv)

Telling an LLM to "be enthusiastic" raises its sycophancy rate from 30% to 50% on a lightly-aligned model, but has zero effect on a strongly-aligned one. We define this gap as the alignment floor, Deltafloor(m)=maxpS(m,p)-minpS(m,p), the range of sycophancy rates a model produces across persona conditions, and treat sycophancy as a persona-conditional property rather than a fixed model property. Pluralistic AI relies on behavioral adaptation via persona prompts like "be creative" or "be thorough", which let systems respect diverse user values and communication styles; the safety question is how much customization a given model can absorb before its truthfulness shifts. We present a controlled case study contrasting a strongly-aligned RLHF + Constitutional-AI model (Claude Sonnet 4.6) with a more lightly-aligned model (Amazon Nova Lite), spanning seven persona conditions and five tasks for 1800 total runs. An existence-pair result motivates per-model auditing: there is at least one strongly-aligned model with Deltafloor=5pp (within 5pp of the 15% control rate) and at least one lightly-aligned model with 45pp (5%--50% range). On the lightly-aligned model, all five Big Five personas increase sycophancy over control, and counterintuitively Agreeableness produces the smallest increase, not the largest. The single largest effect in the study is constructive: a Skeptic persona reduces sycophancy by 25pp on the lightly-aligned model, and is the only persona that instructs resistance against user claims rather than engagement with them, suggesting a directionality account. Cross-model transfer of persona effects is near-zero, so persona-alignment testing must be per-model. We propose Deltafloor as a deployment-time audit metric: measure it on a small persona panel before deploying persona customization.